A completely random dataset, for the purpose of teaching concepts like over fitting and other things. There is no correlation anywhere. Every column is uniformly random.
Dataset Structure
| Attribute | Type | Description |
|---|---|---|
| Random Chemical 1 | Numerical | Random values |
| Random Chemical 2 | Numerical | Random values |
| Random Chemical 3 | Numerical | Random values |
| Random Chemical 4 | Numerical | Random values |
Lab Questions
- Train the DecisionTreeReggresor on any two columns. What is the accuracy? Is the model overfitting? If you add a validator(like KFold) what happens to the accuracy?
- Train the RandomForestRegressor on any two columns. Modify the hyper parameters
n_estimator, change it to 1, 10, and 10000. Which parameter overfits the most?- Train the DecisionTreeReggresor on any two columns. Modify the hyper parameters
max_depth, change it to 3, 5, and 1000. How does this change the model?
Dataset Source
python’s random function.